SPATS: a practical system for comparative analysis of spatio-temporal graph neural networks

Abstract

Thanks to technological advances in sensors and artificial intelligence, large amounts of data that combine spatial and temporal information are being produced in multiple domains. Spatio-temporal graph neural networks (STGNNs) have been recognized as highly effective models for analyzing spatio-temporal data, and so numerous novel STGNN models have recently been developed. However, no systematic and in-depth study has been carried out on the existing STGNN models with various datasets. Thus, it remains to be undecided whether more recent methods achieve better performance than traditional approaches. In this study, we propose a practical system, called SPAtio-Temporal graph System (SPATS), that performs effectively and efficiently the fair comparison of various STGNN models and datasets. SPATS introduces a unified data format to reduce dependency on data models and exploits GPU clusters to handle a large number of model comparisons automatically. Extensive experiments demonstrate that SPATS can efficiently compare STGNN models with reduced memory footprints and fully exploit GPU clusters. Furthermore, SPATS allows us to easily find the effective combination between the STGNN models and the datasets in various domains that have not been examined before.